Valuing crop diversity in biodiesel production plans
Antonella Baglivi1, Giulia Fiorese2*, Giorgio Guariso1, Clara Uggè3
1Dept. of Electronics, Information and Bioengineering, Politecnico di Milano, Milan, Italy
2European Commission, Joint Research Centre, Institute for Environment and Sustainability, Via E. Fermi 2749, I-21027 Ispra (VA), Italy
3ETA-Florence Renewable Energies, Florence, Italy
*Correspondingauthor:GiuliaFiorese, European Commission, Joint Research Centre, Institute for Environment and Sustainability, Via E. Fermi 2749, I-21027 Ispra (VA), Italy
phone number +39 0332 786280
e-mail:
Keywords: Energy crops; Land use; Optimization; Multi-objective nonlinear program; Pareto optimal solutions.
Paper type: Original Research Articles
Abstract
This paper presents a multi-objective modelling framework to define a short-term agricultural plan for biodieselexploitation. The first part of the modelling framework consists in the analysis of local land and climate features in order to evaluate which energy crop can be successfully grown. This phase is performed at local scale using GIS data and software. The second part consists in the formulation of a multi-objective mathematical programming problem. Using the land to be cultivated in each parcel with each crop as decision variables, The objectives of the problem maximize threeareobjectives: the maximization of the net energy produced, of the greenhouse gases avoided with respect to conventional fossil fuels and of the diversity of the energy crop mix. The last is quantitatively measures using a well-known biodiversity index, which allows to study the trade-off between a more varied crop mix and the other two objectives along the frontier of Pareto efficient solutions. The solution of the problem provides the land to be cultivated in each parcel with each crop, the energy required at all stages, the associated greenhouse gases flows and an index of crop diversity. Results allow to define a set of Pareto efficient solutions. It is thus possible to study the trade-offs among the three objectives. The proposed methodologyis applied to aregion of Mato Grosso, Brazil, where biodiesel is produced from oleaginous crops.
Introduction
There is much interest in the production of fuels for the transport sector from biomass resources. Thisis driven by the potential reduction of greenhouse gases (GHG) emissions with respect to traditional fossil fuels and thus by the contribution to global climate change mitigation(Chum et al., 2011). Biofuels may also decreasethe dependency of fuel supply from oil producing countries. Moreover, social and environmental conditions could benefit from improved management of local resources. However, there are also several drawbacks that derive from the production of biofuels(Nature, 2011). First, GHG emission savings should be carefully estimated taking into account all issues which could significantly alter the overall balance, such asindirect land use change(Searchinger et al., 2009),conversion of forest or grassland to produce foodcrop–based biofuels (Fargione et al., 2008), co-products (Farrell et al., 2006) and indirect emissions from fertilizers (Crutzen et al., 2007). Second, land is a limited resource that provides many vital goods and services (e.g., Tilman et al., 2009). The food vs. energy competition for land could spark an increase in food prices. FAO has indeed expressed concerns about the negative effects on food availability and prices due to the expansion of biodiesel production (OECD/FAO, 2012). Furthermore, the extensive cultivation of energy crops to supply large quantity of feedstock to the biomass-to-energy industry could imply extensive monocultures, with consequences on biodiversity (Dale et al., 2010).
There are numerousstudies that deal with the multifaceted nature of biomass-to-energy systems (seeScott et al., 2012 for a review). Harsono et al.(2012), for example, present a cradle-to-gate assessment of the energy balances and GHG emissions of Indonesian palm oil biodiesel production, including all stages of production and land use change. Estimation of the GHG mitigation potential of bioenergy crops requires the consideration of spatially varying information, such as the spatial component of C dynamics under planting of different bioenergy crops (Hillier et al., 2009).Several papers formulate and solve a mathematical problem to optimally plan the bioenergy use in a given area; this type of problem allows determining the optimal type, size and location of conversion plants, the supply area of the plants to give the maximum economic profitability (Čuček et al., 2010)or the maximum net energy(Fiorese et al., 2013).Land competition between bioenergy and traditional food crops can also be included (Andersen et al., 2012). Typically, these studies thus solve a long-term planning problem, (implicitly) assuming that the system to be optimized will remain as it is in the future;for example, assuming that the demand of biofuel is fixed to a given value (Leãoet al., 2011).
Multi-criteria analysis canbe usedto compare different alternatives. For example, Zhou et al. (2007)defined an aggregating function combining indicators on cost, global warming potential, net energy yield and the potential for non-renewable resource depletion. A similar approach has been proposed by Evans et al.(2010) to compare three technologies, while an evolutionary algorithm is used inAyoub et al. (2009)to optimize energy efficiency, total cost, CO2 emissions, or to maximize employed labor hours.
Finally, a more comprehensive formulation of the problem can be obtained with a multi-objective frameworkto simultaneously assess energy, economic, social and environmental impacts and their consequent trade-offs. In a recent survey on 128 papers dealing with energy and environment, Wang and Poh (2014) have categorized 22 as using a multi-objective approach. Only 6 of them, however, were dealing directly or indirectly with biomass exploitation. For example,Pérez-Fortes et al.(2012) formulate a three-objective (economy, environment, society) problem of bio-based electricity generation and describe the Pareto set of solutions, while Tan et al. (2009) used land use, water and carbon footprints as objectives. While in most of these studies the impact on the environment is computed only in terms of GHG emissions, Zhang et al. (2010) have added erosion and N and P losses as environmental objectives of a crop allocation study in Michigan, USA. As suggested by Čuček et al. (2012), other indicators are needed in particular to evaluate the impacts of the bioenergy production chain on biodiversity. Stoms et al. (2012) proposed for instance an approach with a specific emphasis on wildlife aspects, but a quantitative objective to account for crop diversity is still a research theme.
We propose in this paper to use a well-known ecological indicator to explicitly highlight the effects of biodiesel exploitation decisions on crop diversity. This allows to quantitatively evaluate the differences between a variety of energy crops, and the adoption of the single most productive crop, which would entail the diffusion of extensive monoculture, with all the well-known environmental consequences (e.g., Eggers et al., 2009).
Following the approach delineated in the above literature, we thus formulatea multi-objective modelling framework of short-term agricultural plans for biodiesel exploitation at regional scale asa mathematical programming problem with three objectives to be maximized: the net energy produced, the GHG emissions avoided and the diversity of energy crops. The solution of such a problem defines all the alternatives in terms ofthe land to be cultivated each year with each crop in each parcel of the considered domain, the optimal allocation of the harvested crops to thebiodiesel conversion plants, and the energy used in all the process stages under various assumptions about external conditions. The non-dominated alternatives constitute the Pareto boundary of the efficient plans, which is of valuable importance to support the decision making process.
The problem as it is formulated should, in principle, be applied every year to decide what is called the “cropping plan” in agronomic terms (Dury et al., 2012), i.e., the optimal cultivation for the next year (or few years, in the case of perennial crops). Thus, it is inherently different from most of the main current scientific literature(Sharma et al., 2013), where the aim is the long-term optimization of an entire bioenergy supply chain (including the type and location of conversion plants) under the assumption of known demand of products and energy. A further difference is that we assume, as in Zhang et al. (2010), the possibility of substitution between different crops and thus no specific demand for each of them. It is in fact clear that one of the main aims of a biodiesel plan is the maximization of the production independentlyof the feedstock used and that, in general, biodiesel produced in any specific region enters a very large market where it mixes with that of many other areas and also with fossil diesel before reaching the distribution network, where the demand satisfaction finally becomes a constraint.
The introduction of a quantitative way to measure crop diversity allows to trade-off this indicator with energy production thus providing a new way of evaluating the (marginal) environmental “cost” of energy, or, reversely, the energy “cost” of increasing biodiversity. Though in fact crop diversity has been often quoted as a relevant decision criterion (see, for instance the expert survey by Buchholz et al., 2009, or the literature review by Lähtinen et al., 2014), attempts to confront its impact on bioenergy production are still missing.
As a case study, we focus on a specific geographic area in Brazil and on biodiesel from oleaginous crops.
Materials and methods
The modelling framework that we propose is composed of the following fivesteps. The first three, that involve a substantial use of GIS, have been successfully adopted in previous studies in Southern (Fiorese and Guariso, 2010) as well as NorthernEuropean (Tenerelli and Carver, 2102) contexts.
1.Crop selection: the climatic characteristics of the area are analyzed and compared with the phytology and climatic requirements of a set of potentially interesting oleaginous crops.
2.Identification of the suitable area and crop adaptability: through GIS analysis, the area that satisfies the pedological and morphological requirements is identified. Maps of adaptability for each oleaginous crop are created, which quantify the degree of adaptability of the crop to the local characteristics of the territory.
3.Identification of available area: not all the area suitable for agriculture, identified in step 2, is available for oleaginous crops. It can be reduced by environmental, social and political constraints;this results in an often significant reduction of complexity of the original problem as suggested in one of the approaches presented in Lam et al. (2011). Current land uses are therefore analyzed and assumptions are made with respect to the amount of land that can be converted to oleaginous crops.Once this step is accomplished, the area is divided into a number of cells (or parcels) that are small enough to be considered having the same soil and climate characteristics. This is an essential step to convert the continuous territorial representation offered by GIS analysis into a finite number of decision variables, namely what to cultivate in each parcel. Clearly, a finer subdivision allows a very detailed analysis, but on the other hand it may slow down the subsequent phase. As usual, a compromise between detail and speed must be found.When addressing long-term planning problems (see the literature quoted in the introduction), a similar analysis is performed also to identify potential locations for transformation plants.When dealing with annual agricultural planning, on the contrary, the plant location and capacity is fixed and the distances between them and each parcel are easily extracted using standard GIS tools. Given that feedstock has to be transported to the conversion plant, such distances enter in fact both in the energy and in the emission balances.
4. Definition of the study boundary and local peculiarities. The life cycle assessment approach requires that the boundaries of the problem are clearly identified. The farther they are set, the more complete the evaluation of energy and emission balance is. However, it must be stressed that, as a support to decision making, these studies are more aimed to the comparison of alternative plans than to a precise evaluation of all factors. Indeed, if we fix a relative large boundary for the study, we may hope that what has not been considered does not differ too much when varying the possible alternatives. On the other side, even when trying to accurately consider all factors influencing the energy or the emission balances, it is practically impossible to evaluate all the indirect effects that they may imply. Therefore, an accurate analysis of the local situation is essential. While in fact the approach to biodiesel development plans can be general, the optimal decisions may differ in each local situation. Peculiarities range from physical to economical to normative and social settings. For instance, the local merchantability of biodiesel by-products is a relevant factor. Whatever oleaginous crop is processed, a significant part of its biomass and energy remains in various forms after the refining process. The possibility of a local use of these by-products may significantly modify both the energy and the emission balances. Other relevant local factors are crop productivity, farming traditions, availability of resources in the local market, limits in the use of nitrogen fertilizers.
5.Optimal allocation of crops: a multi-objective problem is formulated and solved in order to assess the net energy produced, the GHG emissions avoided and the cropdiversity.
It is important to look at this approach in a correct perspective. The results obtained by any mathematical planning problemare just useful to highlight the most relevant trade-offs, the effect of some scenario variable, the plausible environmental consequences of the decisions being considered. In no way such results must be viewed as the actions to be straightforward adopted. The reason for this is simple: it is evident that real life is always in evolution. The introduction of new crops is an adaptive process: farmers learn progressively the most effective way to cultivate them and thus the productivity changes in time (Uggè et al., 2013). The local market may adapt more or less rapidly to the new products and by-products, but is also influenced by the fluctuations of the national and international markets. The change in crops modifies the carbon balance in the soil and the time to reach a new equilibrium with the new aboveground biomass may be of tens of years (see, for instance,Gelfand et al., 2011). The climatic conditions may differ substantially and for a prolonged period from the average conditions assumed in any planning study and finally such average conditions may never be replicated under future climate changes.
Mathematical formulation of the problem
The three-objective mathematical programming problemcan be formulated as follows.The decision variables zijsfare the fractions of biomass cultivated in cell i, with crop s and method f (for example, in the following case study we distinguish between industrial and family farming) and shipped to plant j for processing. As it is well-known, in long term planning problems, additional variables are needed to determine size (e.g., Andersen et al., 2012), location and technology of conversion plants (e.g., Frombo et al., 2009), which are on the contrary known values in yearly plans.
In the scientific literature(e.g., Leãoet al., 2011), the maximization of an economic objective is often assumed. In this analysis, we chose not to consider any economic variables because of the high variability of their trend, driven by several external factors which we cannot consider here, such as price and demand of different commodities, international trade and so on. For example, Figure 1 shows the evolution of the price index of cereals and of soybeans in Brazil. This index is calculated with respect to the 2004-2006 average prices of the two commodities and provides an indication of their rather large price variability. Additionally, an analysis of both Brazilian and international prices of the last ten years for various crop oils (see for instance, the publications of the Brazilian Ministry of Mines and Energy or of USDA) shows that the prices may considerably vary between different years with a standard deviation between 25 and 30% of the average values. In this paper, we thus consider the energy content of each crop which is indeed a proxy for its economic value, but is fixed and does not depend on the (often unpredictable) market fluctuations as prices would do.
Figure 1
Energy objective
The first objectiveJe, which represents the maximization of the net energy output of the regional biodiesel network, can be written as follows:
(1)
where the first term represents the energy output (at the conversion plants), the second is the energy spent to transport the feedstock to the plants and the last two are the energy employed for the cultivation and the conversion processes.
More precisely,
Nc, Np,Nsand Nf represent, respectively, the number of cells, plants, crops, and farming methods considered;
Af is the land surface (ha) available in each cell for each farming method, that does not depend on the cell index i, when a regular grid is assumed;
E is the energy content of a unit mass of biodiesel (GJ/kg);
ps is the crop productivity obtainable in the best soil conditions in terms of mass of seeds per unit area (kgseeds/ha);
sisrepresents the suitability of cell i for crop s:following the approach inFischer et al. (2010),it may be assumed to vary between 1 (when the area has all the best characteristics needed for crop s) and 0.8 (when not all the best requirements are met).
ws is the amount of biodiesel that can be extracted from a unit weight of seeds type s(kgdiesel/kgseeds). It takes into account both the oil content of the seeds and the efficiency of the extraction operations.
As to the second term,
etris the energy necessary to transport a unit of biomass over a unit distance (GJ/(kg·km));
dij is a measure of the distance between cell i and plant j (km).
Finally, the last term representsthe energy costs of biodiesel production, they are composed by two parts, the first being the energy necessary for the cultivation itself, the second for the transformation of seeds into biodiesel, so:
is the energy of all agricultural operations to cultivate a unit area of crop s (GJ/ha), and